Optical Flow Estimation on KITTI 2015 (test)
3.26Fl-allDDVM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DDVMTemporal Context=Two-frame2026.03 | 3.26 | — | |
| WAFT-DINOv3-a2Temporal Context=Multi-frame2026.03 | 3.56 | — | |
| CrocoFlowTemporal Context=Two-frame2026.03 | 3.64 | — | |
| VideoFlow-MOFTemporal Context=Multi-frame2026.03 | 3.65 | — | |
| MemFlow-TTemporal Context=Multi-frame2026.03 | 3.88 | — | |
| MegaFlowTemporal Context=Multi-frame2026.03 | 3.94 | — | |
| FlowDiffuserTemporal Context=Multi-frame2026.03 | 4.17 | — | |
| StreamFlowTemporal Context=Multi-frame2026.03 | 4.24 | — | |
| SEA-RAFT (L)Temporal Context=Two-frame2026.03 | 4.3 | — | |
| AnyFlowTemporal Context=Multi-frame2026.03 | 4.41 | — | |
| VideoFlow-BOFTemporal Context=Multi-frame2026.03 | 4.44 | — | |
| SAMFlowTemporal Context=Multi-frame2026.03 | 4.49 | — | |
| FlowFormer++Training Data=C+T+S+K+H, tile technique=true2023.03 | 4.52 | — | |
| S-FlowTraining Data=C+T+S+K+H2023.03 | 4.64 | — | |
| RPKNetTemporal Context=Two-frame2026.03 | 4.64 | — | |
| FlowFormerTraining Data=C+T+S+K+H, tile technique=true2023.03 | 4.68 | — | |
| FlowFormerTemporal Context=Two-frame2026.03 | 4.68 | — | |
| GMFlowNetTraining Data=C+T+S+K+H2023.03 | 4.79 | — | |
| CRAFTTraining Data=C+T+S+K+H2023.03 | 4.79 | — | |
| SKFlowTraining Data=C+T+S+K+H, warm-start strategy=true2023.03 | 4.84 | — | |
| RAFT (2-view)Training Data=C + T + S + K + H2020.03 | 5.1 | — | |
| RAFT (warm-start)Training Data=C + T + S + K + H2020.03 | 5.1 | — | |
| RAFTTraining Data=C+T+S+K+H, warm-start strategy=false2023.03 | 5.1 | — | |
| RAFTTraining Data=C+T+S+K+H, warm-start strategy=true2023.03 | 5.1 | — | |
| RAFTTemporal Context=Two-frame2026.03 | 5.1 | — | |
| GMATraining Data=C+T+S+K+H, warm-start strategy=false2023.03 | 5.15 | — | |
| GMATraining Data=C+T+S+K+H, warm-start strategy=true2023.03 | 5.15 | — | |
| GMATemporal Context=Two-frame2026.03 | 5.15 | — | |
| RAFT (2-view)Training Data=C + T + S/K2020.03 | 5.27 | — | |
| RAFT-ftLearning Setting=Supervised, Training Data=S/K, Fine-tuned=true2022.11 | 5.27 | — | |
| MaskFlowNetTraining Data=C + T + S + K + H2020.03 | 6.1 | — | |
| MaskFlowNetTraining Data=C+T+S+K+H2023.03 | 6.1 | — | |
| FM-RAFTTraining Data=C+T+S+K+H2023.03 | 6.17 | — | |
| SciFlowBackbone=MobileFlow, GMACs=78.8, EMA (E)=true, Confidence-base loss masking (C)=true, Semantic cross-domain interference (I)=true, Unlabeled Dataset=WSVD2026.06 | 6.24 | 1.65 | |
| VCNTraining Data=C + T + S + K + H2020.03 | 6.3 | — | |
| RAFT-ftLearning paradigm=Supervised, Fine-tuning=true2020.12 | 6.3 | — | |
| VCNTraining Data=C+T+S+K+H2023.03 | 6.3 | — | |
| HD3Training Data=C + T + S/K2020.03 | 6.55 | — | |
| ScopeFlowTraining Data=C + T + S/K2020.03 | 6.82 | — | |
| IRR-PWCTraining Data=C + T + S/K2020.03 | 7.65 | — | |
| IRR-PWC+ftLearning paradigm=Supervised, Fine-tuning=true2020.12 | 7.65 | — | |
| PWC-Net+Training Data=C + T + S + K + H2020.03 | 7.72 | — | |
| PWC-Net+Training Data=C+T+S+K+H2023.03 | 7.72 | — | |
| LiteFlowNet2Training Data=C + T + S + K + H2020.03 | 7.74 | — | |
| LiteFlowNet2Training Data=C+T+S+K+H2023.03 | 7.74 | — | |
| SciFlowBackbone=RAFT, GMACs=808.9, EMA (E)=true, Confidence-base loss masking (C)=true, Semantic cross-domain interference (I)=true, Unlabeled Dataset=KM2026.06 | 8.4 | 2.27 | |
| MDFlowLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 8.91 | — | |
| SciFlowBackbone=MobileFlow, GMACs=78.8, EMA (E)=true, Confidence-base loss masking (C)=true, Semantic cross-domain interference (I)=true, Unlabeled Dataset=KM2026.06 | 8.95 | 2.78 | |
| GMFlowTraining Data=C+T+S+K+H2023.03 | 9.32 | — | |
| LiteFlowNetFine-tuning=true2020.06 | 9.38 | — | |
| FDFlowNetFine-tuning=true2020.06 | 9.38 | — | |
| LiteFlowNet-ftFine-tuning status=Yes, Params (M)=5.37, FLOPS (G)=163.5, Time (s) 1080Ti=0.055, Time (s) TX2=0.9072021.03 | 9.38 | — | |
| UPFlowLearning paradigm=Unsupervised2020.12 | 9.38 | — | |
| UPFlowLearning Setting=Unsupervised, Training Data=S/Kraw2022.11 | 9.38 | — | |
| PWC-NetFine-tuning=true2020.06 | 9.6 | — | |
| PWC-Net-ftFine-tuning status=Yes, Params (M)=8.75, FLOPS (G)=90.8, Time (s) 1080Ti=0.034, Time (s) TX2=0.4852021.03 | 9.6 | — | |
| PWC-Net+ftLearning paradigm=Supervised, Fine-tuning=true2020.12 | 9.6 | — | |
| PWC-Net-ftLearning Setting=Supervised, Training Data=S/K, Fine-tuned=true2022.11 | 9.6 | — | |
| PWC-NetTemporal Context=Two-frame2026.03 | 9.6 | — | |
| ASFlowLearning Setting=Unsupervised, Training Data=S/Kraw2022.11 | 9.67 | — | |
| PATS + Intp.Training Data=M (Megadepth), Interpolation=true2023.03 | 9.68 | 3.39 | |
| FLCLearning Setting=Unsupervised, Training Data=KVO + K(Stereo), Stereo=true2022.11 | 9.7 | — | |
| SciFlowBackbone=RAFT, GMACs=808.9, EMA (E)=true, Confidence-base loss masking (C)=true, Unlabeled Dataset=KM2026.06 | 9.7 | 2.81 | |
| SciFlowBackbone=MobileFlow, GMACs=78.8, EMA (E)=true, Confidence-base loss masking (C)=true, Unlabeled Dataset=KM2026.06 | 9.7 | 3.24 | |
| OIFlowLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 9.81 | — | |
| SciFlowBackbone=RAFT, GMACs=808.9, EMA (E)=true, Unlabeled Dataset=KM2026.06 | 10.1 | 2.85 | |
| LiteFlowNet+ftLearning paradigm=Supervised, Fine-tuning=true2020.12 | 10.24 | — | |
| CoT-AMFlowLearning Setting=Unsupervised, Training Data=Sraw/Kraw2022.11 | 10.34 | — | |
| FlowNet2-ftLearning Setting=Supervised, Training Data=S/K, Fine-tuned=true2022.11 | 10.41 | — | |
| DistillFlowLearning Setting=Unsupervised, Training Data=S + Sraw/Kraw2022.11 | 10.54 | — | |
| Flow2StereoLearning Setting=Unsupervised, Training Data=K(Stereo), Stereo=true2022.11 | 11.1 | — | |
| UFlowLearning paradigm=Unsupervised2020.12 | 11.13 | — | |
| UFlowLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 11.13 | — | |
| FastFlowNet-ftFine-tuning status=Yes, Params (M)=1.37, FLOPS (G)=12.2, Time (s) 1080Ti=0.011, Time (s) TX2=0.1762021.03 | 11.22 | — | |
| FastFlowNet-ftLearning Setting=Supervised, Training Data=S/K, Fine-tuned=true2022.11 | 11.22 | — | |
| MDFlow-FastLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 11.43 | — | |
| FlowNet2Fine-tuning=true2020.06 | 11.48 | — | |
| FlowNet2Training Data=C + T + S/K2020.03 | 11.48 | — | |
| FlowNet2-ftFine-tuning status=Yes, Params (M)=162.52, FLOPS (G)=24836.4, Time (s) 1080Ti=0.116, Time (s) TX2=1.5472021.03 | 11.48 | — | |
| DistractFlowBackbone=RAFT, GMACs=808.9, EMA (E)=true, Unlabeled Dataset=Sintel+KM2026.06 | 11.7 | 3.01 | |
| ARFlowLearning paradigm=Unsupervised2020.12 | 11.8 | — | |
| ARFlowLearning Setting=Unsupervised, Training Data=Sraw/Kraw2022.11 | 11.8 | — | |
| ECO-TR + Intp.Training Data=M (Megadepth), Interpolation=true2023.03 | 12.1 | 3.16 | |
| MRFlowTraining Data=S2020.03 | 12.19 | — | |
| PDC-Net+Training Data=M (Megadepth)2023.03 | 12.62 | 4.53 | |
| SimFlowLearning paradigm=Unsupervised2020.12 | 13.38 | — | |
| SimFlowLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 13.38 | — | |
| COTR + Intp.Training Data=M (Megadepth), Interpolation=true2023.03 | 13.65 | 3.65 | |
| STFlowLearning paradigm=Unsupervised2020.12 | 13.83 | — | |
| STFlowLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 13.83 | — | |
| SelFlowLearning paradigm=Unsupervised2020.12 | 14.19 | — | |
| SelFlowLearning Setting=Unsupervised, Training Data=Sraw/K2022.11 | 14.19 | — | |
| DDFlowLearning paradigm=Unsupervised2020.12 | 14.29 | — | |
| DDFlowLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 14.29 | — | |
| FlowFormerTraining Data=C + T (FlyingChairs + FlyingThings)2023.03 | 14.72 | 4.09 | |
| FlowFields++2020.03 | 14.82 | — | |
| DCFlowTraining Data=S2020.03 | 14.86 | — | |
| FlowFields2020.03 | 15.31 | — | |
| EpiFlowLearning paradigm=Unsupervised2020.12 | 16.95 | — | |
| EPIFlowLearning Setting=Unsupervised, Training Data=C + S/K2022.11 | 16.95 | — |